Edge Delamination of Monolayer Transition Metal Dichalcogenides

Jul 11, 2017 - Delamination of thin films from the supportive substrates is a critical issue within the thin film industry. The emergent two-dimension...
2 downloads 0 Views 7MB Size
Download PDF
Edge Delamination of Monolayer Transition Metal Dichalcogenides Thuc Hue Ly,*,†,‡ Seok Joon Yun,# Quoc Huy Thi,# and Jiong Zhao*,† †

Department of Applied Physics, The Hong Kong Polytechnic University, Hong Kong SAR, People’s Republic of China Department of Chemistry and Center of Super-Diamond & Advanced Films (COSDAF), City University of Hong Kong, Hong Kong SAR, People’s Republic of China # IBS Center for Integrated Nanostructure Physics, Institute for Basic Science, Sungkyunkwan University, Suwon 440-746, Korea ‡

S Supporting Information *

ABSTRACT: Delamination of thin films from the supportive substrates is a critical issue within the thin film industry. The emergent two-dimensional, atomic layered materials, including transition metal dichalcogenides, are highly flexible; thus buckles and wrinkles can be easily generated and play vital roles in the corresponding physical properties. Here we introduce one kind of patterned buckling behavior caused by the delamination from a substrate initiated at the edges of the chemical vapor deposition synthesized monolayer transition metal dichalcogenides, led by thermal expansion mismatch. The atomic force microscopy and optical characterizations clearly showed the puckered structures associated with the strain, whereas the transmission electron microscopy revealed the special sawtooth-shaped edges, which break the geometrical symmetry for the buckling behavior of hexagonal samples. The condition of the edge delamination is in accordance with the fracture behavior of thin film interfaces. This edge delamination and buckling process is universal for most ultrathin two-dimensional materials, which requires more attention in various future applications. KEYWORDS: tungsten disulfide, molybdenum disulfide, edge delaminate, AFM, TEM, PL According to the thin film technology, the interfaces between deposited films and substrates occasionally have large strains, including lattice mismatch strain,16 thermal mismatch strain,17 or some other external loading induced strains.18 The lattice mismatch strain often occurs in the metallic or covalent bonding interfaces, whereas for van der Waals (vdW) 2D materials the thermal mismatch strain or external loading strain is prevalent. All the chemical vapor deposition (CVD)synthesized 2D TMD specimens need to be cooled from high temperature to room temperature for the completion of the synthesis.19,20 Additionally, 2D materials normally have distinct in-plane thermal expansion coefficients (CTEs) compared to those of the substrates,21 and considerable thermal mismatch strain can be expected between the 2D materials and the substrates. Therefore, the 2D materials we are studying are not the “as-grown” samples but postgrowth strained samples.

he emergent two-dimensional (2D) materials1−3 consist of only single or a few atomic layers in thickness. Despite that well-established models of 2D materials are atomically flat,4 some experiments have also found noticeable out-of-plane fluctuations, especially in free-standing 2D materials.5,6 Basically these 3D fluctuations, which break the long-range order, should be responsible for their unexpected ultrahigh stability.7 In another aspect, for those that are well supported by rigid substrates, 2D materials are expected to fully comply with the surface morphology of the underneath substrates due to intrinsic flexibility,8,9 and owning to this flexibility, wrinkles or buckles in 2D materials can be readily induced.8−11 It has been demonstrated that the interfacial coupling between atomistic layers can play deterministic roles in their physical properties.12,13 Herein we intend to address the edge effect on the buckling and wrinkling of 2D materials, more specifically, in transition metal dichalcogenides (TMDs). In addition, our experimental characterizations and analysis of monolayer TMDs clarified the previous substantial misunderstandings in “domain structures”14 or “edge enhancement”15 associated with their optical or chemical properties. The origin of the domain contrast is probably the geometrical buckling in the flakes, rather than the atomic defects.14,15

T

© 2017 American Chemical Society

Received: June 19, 2017 Accepted: July 11, 2017 Published: July 11, 2017 7534

DOI: 10.1021/acsnano.7b04287 ACS Nano 2017, 11, 7534−7541

Article

www.acsnano.org

Article

ACS Nano

Figure 1. Schematic and SEM characterizations of the WS2 flakes on SiO2 substrates. (a) Illustration of the hexagonal buckled monolayer WS2. The in-plane tension is highlighted by arrows. The two insets show the atomic structures of two different type edges, nonbuckled S-ZZ edge and buckled W-ZZ edge. (b) Secondary electron images of the as-synthesized WS2 flakes. Scale bar is 20 μm. (c) Higher magnification secondary electron image of the “six patch” domain contrast in a hexagonal WS2 monolayer after transfer to a new silicon substrate. The white dot in the center is a marker on the silicon wafer. Scale bar is 10 μm.

Figure 2. AFM characterizations of the WS2 monolayers. (a) Contact mode (CM) topographic AFM image of as-synthesized monolayer WS2. The inset shows the height profile across the edge. (b) Tapping mode (TM) topographic AFM image of monolayer WS2 with domain contrast after transfer. (c) Friction force microscopy (FFM) image of as-synthesized monolayer WS2 with domain contrast. All scale bars are 10 μm.

implying the nonuniform morphology of hexagonal WS2 flakes. As the next step, the as-grown samples were checked by atomic force microscopy (AFM) experiments. Figure 2a,b show the AFM topographic images in contact mode (AFM-CM) and tapping mode (AFM-TM), respectively. The monolayer WS2 has a thickness of around 0.8 nm (Figure 2a). The topographic image indicates the as-grown WS2 flake is flat when the AFM tip is forced to push the sample (in contact mode) (Figure 2a); however the height contrast in the domains emerges by the noncontact mode (tapping mode) (Figure 2b). This is direct evidence of flake buckling in the brighter domains in Figure 2b. The friction force microscopy (FFM) of the as-grown WS2 hexagonal flake is shown in Figure 2c. Unsurprisingly, the FFM image shows similar domain contrast to the SEM SE and AFMTM images in the hexagonal flakes. Note that the edges of the hexagons can be divided into two types (Figure 1a), W-zigzag (W-ZZ) or S-zigzag (S-ZZ).22 The edge types of the domains can be identified by transmission electron microscopy (TEM), using two methods: selected area diffraction pattern (SAED) and high-resolution scanning transmission electron microscopy (STEM) (refer to Supporting Information Figures S3, S4). TEM results confirmed dark (lower) contrast domains in SEM SE micrographs (Figure 1c), buckled domains in AFM-TM (Figure 2b), and high-friction domains in FFM (Figure 2c), all corresponding to W-ZZ edge domains. Further, high spatial resolution photoluminescence (PL) mapping on the WS2 flakes (Figure 3) again exhibits the

RESULTS AND DISCUSSION Our WS2 monolayers were grown on Si/SiO2 substrates via CVD methods;22 specific growth details can be found in the Methods section. The optical microscopy (Supporting Information Figure S1) and scanning electron microscopy (SEM) demonstrate the morphologies of the WS2 flakes (Figure 1). Similar to all the previous literature23−26 and comparable with other TMDs in the family (MoS2, MoSe2, WSe2), the monolayer WS2 has well-faceted edges, which preserves the 3-fold symmetry of the crystal structure (point group: D3h). Some of them are equatorial triangles, while others have truncated triangle shapes or are even close to equatorial hexagons. Some previous reports have already discussed the shape evolution for these TMD monolayers during growth,27 and we have prepared another manuscript on the growth mechanisms; hence we focus on the postgrowth mechanical responses in this study. Interestingly, the SEM secondary electron (SE) images of WS2 hexagonal flakes show shallow contrast differences between the nearby domains, even after a one-time poly(methyl methacrylate) (PMMA) transfer process28 to the new Si substrate (Figure 1c). The 3-fold symmetry of the contrast matches well with the crystal symmetry, regardless of the 6-fold symmetry in the hexagonal shape. In addition, triangle WS2 flakes do have similar contrast inhomogeneity within single flakes (Supporting Information Figure S2). The SEM SE imaging is sensitive to the sample surface and charges but insensitive to atomic defects or chemical compositions,29 7535

DOI: 10.1021/acsnano.7b04287 ACS Nano 2017, 11, 7534−7541

Article

ACS Nano

Figure 3. PL characterizations of monolayer WS2 flakes. (a, b) PL intensity mapping image for the as-synthesized hexagonal and quasi-triangle monolayer WS2 flakes. Inset shows the corresponding PL peak position mapping, with wavelength scale bars on the right side. Scale bar of (a) is 7 μm, and (b) is 12 μm. (c) Corresponding PL spectra for the W-ZZ edge domain and S-ZZ edge domain in (a). (d, e) PL intensity mapping image for two WS2 flakes that are transferred to new silicon substrates. Inset shows the corresponding PL spectra peak position mapping; wavelength scale bars on the right side. Scale bar in (d) is 6 μm, (e) is 12 μm. (f) PL spectra for the transferred WS2 sample in different domains in (e).

ZZ domains, implying the release of tensile strain during transfer. In Figure 3c, the PL peak for the W-ZZ edge domain is blueshifted compared with the S-ZZ edge domain. In addition, both the W-ZZ edge domain and S-ZZ edge domain have red-shifted PL peaks compared to the unstrained (suspended or aftertransfer) sample. This is because tensile strain existed in the assynthesized sample, and it is larger in the S-ZZ edge domain than in the buckled W-ZZ edge domain. After transfer, the PL peaks for the W-ZZ domain and S-ZZ domain both blue-shifted due to the relaxation of strain, and the W-ZZ domain has a PL peak with a larger wavelength compared to the S-ZZ domain, possibly because more sulfur vacancies were introduced into the buckled W-ZZ domain than the flat S-ZZ domain during the transfer process. To learn more about the puckered structure, we employed TEM to examine the WS2 monolayers that were transferred onto a Quantafoil TEM grid via the PMMA method.28 Figure 4 presents TEM images of the monolayer hexagonal WS2 sample at both low and medium magnification. The results show that W-ZZ edges are straight and atomically smooth (Figure 4d and Supporting Information Figure S6), while the S-ZZ edges become sawtooth-like in the mesoscale with the teeth period of 50−200 nm (Figure 4c). The edge morphologies mainly result from the difference in surface energies. The W-ZZ edges are more stable with lower formation energy, and the S-ZZ edges preferred to be decomposed into W-ZZ edge sections, forming the 60° sawtooth edges (Figure 4c). The high-resolution annular dark field (ADF) images and diffraction patterns are utilized to analyze the crystal directions; therefore the S-ZZ edges and W-ZZ edges are exclusively determined (Supporting Information Figures S3, S4). In addition, the dark field (DF) TEM images show highdensity wrinkles on the WS2 samples. The DF TEM technique exhibits diffraction contrast, hence is very sensitive to the crystal

domain contrast as in the SEM and AFM results. There is distinct contrast in the full-range integrated PL intensity for both the as-grown sample (Figure 3a,b) and after-transfer samples to new Si wafers (Figure 3e). W-ZZ edge domains all have much higher PL intensities than S-ZZ domains. Moreover, the A exciton peaks in the W-ZZ edge domains have a ∼2 nm blue shift compared to S-ZZ edge domains in the original sample (Figure 3c). In contrast, a red shift of the A exciton peaks was observed in the after-transfer sample (Figure 3f). The inset mapping images of the PL peak positions (Figure 3a,b,e) clearly revealed this opposite trend of peak shift for the W-ZZ and S-ZZ domains. In a recent publication14 the PL contrast in different domains was explained by the different types of atomic defects. However, since the entire flake experiences the same growth history, it is unlikely that, during growth, the diffusion of defects across the domain boundaries can be prohibited and abrupt boundaries in defect distributions are formed or even lead to opposite defect types (cation/anion defects). In our view, the enhancement of PL in the W-ZZ edges can be more reasonably explained by the buckling of W-ZZ domains; thus the suspended flakes can have higher PL abilities than the substrate-supported ones.30 Moreover, the PL intensity is remarkably enhanced on the edge of WS2 (Figure 3a,b,e and Supporting Information Figure S5), which was previously attributed to the higher defect concentration at the edges.15 The flake edge delamination from the substrate can reproduce such PL behavior as well. Similar domain contrast including the edge effects is universally discovered in a lot of TMD materials such as MoS2 and WSe2.31−33 It is also noted that the contrast between domains cannot be observed when the sample size gets smaller ( MoS2 at 750 °C),39 after cooling the tensile strain remaining in the WS2 flake is relatively larger than that of MoS2. The shape and size of the flakes are important because they are correlated with the delamination process of the flakes from the substrate. Delamination and buckling of the thin films with respect to the substrates under tensile stress (σ)40 can be understood by existing theories.41,42 Delamination or buckling are initialized at the edges (Figure 6). The simplest model to describe this is to consider a mode II cracking model under shear (Figure 6b). The substrate exerts shear stress on the edges of the TMD monolayer, which induces propagation of cracks. The critical crack size (c) along the edge has the relationship with the interfacial toughness (Kic) between the film and substrate:43 c=

between suspended 2D materials and substrates. The domain contrast in the AFM-TM image and FFM image (Figure 2) can be straightforwardly explained by the buckles in the W-ZZ domain and subsequent larger friction force on the AFM tip (Figure 2c), similar to the previous analysis of friction origin on graphene.44 The buckles alter the local PL properties such as PL intensity and peak position.45 In our case, the difference of PL intensity between the nonbuckled part and buckled part can be rationalized by the enhanced PL ability for suspended TMD monolayers than the supported ones,29 which comes from the reduced doping effect in suspended membranes. The PL peaks blue shift in the after-transfer samples compared to withouttransfer samples (Figure 3), in agreement with the release of significant thermal mismatch tensile strains. In the second mechanical process, the deposition of PMMA upon the buckling parts in WS2 will introduce wrinkles as shown in Figure 5. In the W-ZZ buckled domains, wrinkles parallel as well as perpendicular to the edges can be formed. However, in the flat S-ZZ domains, vertical wrinkles can be precluded because of the restriction of circumferential deformation; however the free relaxation of the edges in the radial direction can naturally form transverse wrinkles. It should be noted that before transfer of WS2 onto the TEM grid, the changes from buckles into wrinkles may already occur in some samples, and the buckles before transfer sometimes can be fully retained even after the transfer process.14 Some factors such as sample quality and conditions of transfer matter. Moreover, the environmental attack on the WS2 samples such as light and humidity can introduce defects46,47 and significantly reduce the strain and flexibility of the flakes, hence suppressing the delamination and puckering. The degree of delamination is also dictated by the interaction between the substrate and 2D membrane; that is, on the elastomer substrate, the TMD layers can be continuously elongated until 16% tensile strain without relaxation.48,49

K ic πσ 2

(1)

with the initial crack size larger than c along the edge. It is able to catastrophically propagate toward the inner part of the flake and then lead to buckling. If the crack starts to propagate, the crack can continue in the radial direction (Figure 6c) because of the ease of tensile strain release perpendicular to the free edge, whereas in the circumferential direction the tensile strain cannot be released. Therefore the crack will cease if the crack line gradually turns to the radial direction (Figure 6c). The straight W-ZZ edges (Figure 4d) can have larger initial cracks, which is over c compared to the sawtooth-like S-ZZ edge (Figure 4c), as on the sawtooth edge the initial crack size is interrupted by the periodical sawtooth shape, with the period (∼50 nm) as the upper limit of the initial crack size. This explains why the delamination always occurs first in the straight W-ZZ edges and tend to follow 3-fold symmetry for the whole buckling process. The size of the flakes also matters, as larger flakes have a higher possibility to have a larger initial crack and then trigger the delamination. On the contrary, small flakes without a sufficiently large initial crack with a size over c usually have fewer buckles. Furthermore, in WS2 it is easier to find such domain patterns than in MoS2 (Supporting Information Figure S9), in agreement with the greater thermal expansion mismatch strain (σ) at the interface and smaller critical crack size (c) for WS2 delamination. The lower SEM SE contrast for the buckled parts (Figure 1) may be due to the reduction of charge scattering at the interface

CONCLUSIONS In summary, there are many open questions in the inhomogeneity of strain distribution and related physical properties in 2D materials. In this work we have addressed the importance of the edge delamination and the subsequent puckering such as buckles and wrinkles in single atomic layers. The “domain contrast” and “edge enhancement” in 2D materials should involve considerations of the buckling 7538

DOI: 10.1021/acsnano.7b04287 ACS Nano 2017, 11, 7534−7541

Article

ACS Nano

Scanning Electron Microscopy. Field-emission scanning electron microscopy (JSM7000F, Jeol, Japan) was used to examine the surface morphology of samples at different accelerating voltages to obtain a high level of contrast at different magnifications. An accelerating voltage of 10 kV was used to obtain sufficiently pronounced signals while retaining sensitivity to the sample surface. TEM Sample Preparation. The CVD tungsten disulfide was transferred on a hole with a 1.2 μm in diameter in a Cu Quantifoil TEM grid (product no. 658-200-CU) by a PMMA-assistant method. Thin layer PMMA was spin-coated on an as-grown WS2/SiO2/Si substrate (2000 rpm, 1 min). The WS2 and PMMA support were then detached from the SiO2/Si substrate by floating the PMMA/WS2/ SiO2/Si, with the PMMA side up, in a 1 M HF solution. Next, the PMMA/WS2 was washed with deionized water. The PMMA/WS2 layer is scooped out in pieces onto the TEM grid; then PMMA was removed gently by evaporated acetone (acetone was heated to 130 °C), leaving WS2 suspended freely on holes in the TEM grid substrate. Finally, the sample was annealed at 180 °C at high vacuum (10−6 Torr) for 12 h to further remove PMMA. TEM Measurement. TEM experiments were carried out using a JEM ARM 200F machine under 80 kV. The acquisition time for dark field imaging was 1 s using the smallest objective lens aperture, and the reflex (10−10) was always selected for DF imaging. The HR-TEM imaging acquisition time was also 1 s. ADF-STEM. ADF-STEM imaging was conducted with a CEOS aberration-corrector on the same TEM. High-angle annular dark field (HAADF) images were acquired at a 20 mrad convergence angle.

processes associated with the edges. The atomic defects may not be the dominant reason, however; they could be the consequences of mechanical buckling, because the flat and buckled parts have different chemical activities, thereby inducing defects to different degrees in ambient conditions. Overall, herein we have addressed the delamination and buckling issues that can play a vital role in the synthesis as well as in the optical, electrical, or chemical applications of these emergent 2D materials.

METHODS Synthesis of Tungsten Disulfide (WS2) on a SiO2/Si Wafer. WS2 was grown on a SiO2/Si wafer by an atmospheric CVD process. For synthesizing monolayer WS2, we first coated a precursor solution on a SiO2/Si substrate by the spin-casting method. Preparation of the precursor solution was conducted by mixing three types of water-based solutions (the three types of solution are defined as A, B, and C), where A (tungsten precursor) includes 0.1 g of ammonium metatungstate hydrate [(NH4)6H2W12O40·xH2O: Sigma-Aldrich, 463922] dissolved in 10 mL of DI water, B (promoter) contains sodium cholate hydrate (Sigma-Aldrich, C6445) as a promoter dissolved in DI water (0.3 g of SC in 10 mL of DI water), and C (medium solution) is a medium to mix the promoter and precursor, which was prepared from an OptiPrep density gradient medium (Sigma-Aldrich, D1556, 60% (w/v) solution of iodixanol in water). C does not affect the growth but is only for a better spin-casting process. A, B, and C solutions were mixed in a certain ratio for their purpose (discussed below). Then, the mixed solution was coated onto a SiO2/ Si wafer by spin-casting at 3000 rpm for 1 min. A two-zone CVD system was introduced for controlling sulfur and substrate zone temperatures separately. Here, 0.2 g of sulfur (Sigma, 344621) was loaded, while the solution-coated substrate containing the metal precursor was placed in another zone. Synthesis of WS2 in this work was carried out at atmospheric pressure. For growth, the sulfur zone was heated to 210 °C at a rate of 50 °C/min; at the same time, the substrate zone was set to 780 °C. Nitrogen (600 sccm) and 5−20 sccm of hydrogen gas were introduced as carrier gas and reactive agent. Shape Controlling of CVD-Grown WS2 on a SiO2/Si Wafer. For synthesizing triangular-shaped WS2, a precursor solution at a ratio of 1:6:1 was spin-casted on a SiO2/Si wafer. When the substrate temperature reached a maximum (800 °C), 5 sccm of hydrogen was introduced for 10 min. The key factor to control the shape of WS2 is the hydrogen injection timing. Hydrogen accelerates the growth rate of WS2 by enhancing the reduction rate of tungsten oxide. It also affects the etching process. For the hexagonal-shaped WS2 case, the precursor solution ratio is set to 2:6:1 and 10 sccm of hydrogen was introduced from the beginning. Photoluminescent and Raman Spectroscopy. PL and Raman mapping (NT-MDT, 532 nm wavelength, NTEGRA Spectra PNL, 100× lens, 0.7 NA) were performed using a laser (532 nm) with ∼30 μW power. The scanned image was obtained at 128 × 128 pixels with a grating of 1800 g/mm to yield a spatial resolution of 200 nm for confocal PL and 600 g/mm to yield a spectral resolution of